use trait upcasting and drop the as_any/as_any_mut boilerplate

[a10y]

Since Rust 1.86 (which is before our MSRV) there has been builtin trait upcasting, meaning instead of forcing all our traits to have a dummy fn as_any(&self) -> &dyn Any methods with trivial impls, we can instead just take a pointer to the trait object and do as &dyn Any.

This reduces the amount of boilerplate in trait impls, which is nice. Another benefit is that in the past blanket impls like SessionVar would end up getting pulled in for dispatch of calls to as_any() which has caused me frustration. Now the dispatch is clearer.

This isn't a breaking change b/c it doesn't touch the VTables, which is the implementer surface area

[codspeed-hq]

Merging this PR will degrade performance by 75.14%

❌ 9 regressed benchmarks
✅ 384 untouched benchmarks
⏩ 2235 skipped benchmarks¹

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	take_indices[(100000, 1024)]	246.4 µs	970.8 µs	-74.62%
❌	Simulation	take_indices[(100000, 16)]	747.1 µs	1,409.5 µs	-47%
❌	Simulation	take_indices[(10000, 256)]	68.8 µs	139.2 µs	-50.55%
❌	Simulation	take_indices[(100000, 4)]	1.7 ms	2.2 ms	-26.01%
❌	Simulation	take_indices[(10000, 4)]	207 µs	261.3 µs	-20.78%
❌	Simulation	take_indices[(100000, 256)]	268.1 µs	989.1 µs	-72.89%
❌	Simulation	take_indices[(10000, 16)]	116.3 µs	179.1 µs	-35.1%
❌	Simulation	take_indices[(10000, 1024)]	66.4 µs	137.5 µs	-51.67%
❌	Simulation	take_indices[(100000, 4096)]	244.6 µs	984.3 µs	-75.14%

---

_{Comparing aduffy/asany (606ab92) with develop (5ce08ed)}

Footnotes

1. 2235 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[robert3005]

it's interesting that this had changed execution of take benchmarks?